Tile-based rendering can significantly optimize bandwidth by disabling specific cache entries from be evicted to memory. However, graphics drivers do not always know whether these cache evictions can be inhibited when it submits the Graphics Processor Unit (GPU) instructions. For example, some solutions require the graphics driver to specify whether the cache evictions should occur before it queues the GPU commands. When the driver does not have enough information, it must assume the worst-case and always evict the cache to memory. This increases the memory bandwidth and power consumption of the GPU.